Lead ML Research Engineer, Multimodal Driving Models

Fruition Group US; San Francisco (CA); On-site; USD 200,000 - 260,000; Full time

Job summary

Fruition Group US is partnering with a fast-growing autonomous vehicle company to hire a Staff Machine Learning Research Engineer. You will work on large multimodal models that interpret what the vehicle sees and decide next actions, with end-to-end involvement from dataset shaping to onboard deployment.

The role emphasizes taking models from experiments to real-world tests on vehicles, with options to lead a major research thread, implement training pipelines, and shrink models to fit onboard

Qualifications

• 5+ years building and training deep learning models in PyTorch, TensorFlow, or JAX.
• Experience taking models from training to production.
• Strong knowledge of multimodal models, transformers, diffusion, and flow matching.
• Experience with large-scale or distributed training.
• Background with camera, lidar, radar data; publications or autonomous driving experience is a plus.

Responsibilities

• Build, train, and test vision-language-action models for driving decisions.
• Lead a major research thread, from data and design to large-scale training.
• Create training/evaluation pipelines with rigorous driving-scenario metrics.
• Develop model compression to run onboard on vehicle hardware.
• Apply supervised fine-tuning and RL to improve reasoning and reliability.
• Collaborate with perception, planning, and platform teams to productionize research.

Skills

Deep learning frameworks (PyTorch, TF,

Education

Master's degree in CS/EE/Math/Stats or related field

PhD in related field preferred

Tools

ONNX

TensorRT

Job description

Fruition Group US is partnering with a fast-growing autonomous vehicle company to hire a Staff Machine Learning Research Engineer. You will work on large multimodal models that interpret what the vehicle sees and decide next actions, with end-to-end involvement from dataset shaping to onboard deployment.

The role emphasizes taking models from experiments to real-world tests on vehicles, with options to lead a major research thread, implement training pipelines, and shrink models to fit onboard